gogoymh/auto-ontology-mcp

auto-ontology-mcp brings the semantic layer of NVIDIA Auto Ontology into any MCP client, doing two things: first, it builds business terminology against your database and reviews table associations, with the model in the client writing SQL that the server validates against the ontology and actual data before executing it in read-only mode; second, it serves as a personal knowledge base, storing facts from receipts, bills, papers, and reading notes as summable and comparable records, and storing opinions as concept pages with typed links, each retaining its source and exportable to an Obsidian vault. All steps requiring language capabilities are handled by the model you are currently using, and the server requires no model interface or API Key.

auto-ontology-mcp brings the semantic layer of NVIDIA Auto Ontology into any MCP client, doing two things: first, it builds business terminology against your database and reviews table associations, with the model in the client writing SQL that the server validates against the ontology and actual data before executing it in read-only mode; second, it serves as a personal knowledge base, storing facts from receipts, bills, papers, and reading notes as summable and comparable records, and storing opinions as concept pages with typed links, each retaining its source and exportable to an Obsidian vault. All steps requiring language capabilities are handled by the model you are currently using, and the server requires no model interface or API Key.

Why it matters Text-to-SQL often silently miscalculates; this tool intercepts incorrect associations, flags case sensitivity in enum values, and prevents duplicate counting in one-to-many relationships.

Stars 316 (+51 / 24h)

Reasons

  • 24 hours added 51 stars (indicating attention only)
  • The README uses a comparison table to list typical errors when models read table structures directly, and how each is intercepted after adding the ontology

Usage

  • After installing uv, run uvx --from git+https://github.com/gogoymh/auto-ontology-mcp auto-ontology-mcp demo to create a demo database, then follow the README to integrate with Cursor, VS Code, or Claude apps

Risks

  • Personal project; the README states no affiliation with NVIDIA
  • Effectiveness depends on the capabilities of the client model; confirm read-only permissions before connecting to production databases

Open

GitHub

Explore the latest AI Radar updates